Moritz Schauer

You may have to extend transport by creation and deletion of mass to make sense of weighted measures

I checked with https://en.wikipedia.org/wiki/Conjugate_prior#Continuous_distributions, so I agree that this is correct. Is there a way that we could interchange last Cholesky und inversion and return a Wishart which is parametrized...

Great! This is the original implementation: https://github.com/JuliaStats/Distributions.jl/pull/140/files Maybe @nfoti remembers

Don't worry. Can we cook up a test which shows that this is right and the version on master is wrong?

One suggestion would be to use in test that the one-dimensional case falls back to the InverseGamma/Gamma distribution, so they should give corresponding results.

What is the status here?

@johnczito Would this fall into your area of expertise, be interesting for you?

Great. I also need it as a dependence and can help to keep an eye on it.

It was deleted recently from Distributions, https://github.com/JuliaStats/Distributions.jl/commit/8ed322895dd9e80033dd69a6c4dc21c8e3371f81, see the deleted file in the diff.

Nobody is working on it, but it's not abandoned. Bringing it up to date might not be much work, it's missing some latest version bumps on `SpecialFunctons`, `Distributions` and `PDMats`....